Talent skill optimization path planning method and system based on dynamic man-post matching

Abstract

The invention provides a talent skill optimization path planning method and system based on dynamic man-post matching, and relates to the technical field of intelligent talent matching and optimizing.The method comprises the steps that data physical mapping and inertial index introduction are conducted on obtained multi-dimensional data, and an employee skill state space model is constructed; based on the employee skill state space model, constructing a skill topological graph with skills as nodes and association degrees between the skills as connection edge weights, and obtaining a multi-scale energy network through integration; performing complementation optimization on the skill correlation degree and the employee potential value in the multi-scale energy network through the improved information entropy diffusion model to obtain an optimized multi-scale energy network; and performing multi-target particle game optimization by using the optimized multi-scale energy network and post demand data to obtain a global optimal talent skill optimization path. According to the method, the accuracy of a traditional person and post matching technology for the emerging technology field and the reasonability of an emerging technology talent skill optimization path planning method are greatly improved.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of talent intelligent matching and optimization, in particular to a talent skill optimization path planning method and system based on dynamic person-post matching. BACKGROUND

[0002] With the acceleration of digital optimization process and the upgrading of industrial structure, the demand for emerging skill talents in the employment field is increasing, and emerging skill talents need to continuously update their knowledge and skills to adapt to market changes, which makes the traditional person-post matching accuracy and talent skill optimization path planning method face many challenges.

[0003] The existing person-post matching system technical solution mainly relies on static data and quantitative analysis, however, the updating of talent skills in the new era puts forward more requirements, making it difficult for traditional technology to accurately predict future talent skill demand; and the blurring of job boundaries brought about by digital optimization makes job demand more complex, making it difficult to accurately reflect the dynamic changes in market demand and talent skill demand. The limitations of the existing technology are described below from three aspects: I. Static data and dynamic market demand are out of sync Traditional person-post matching systems usually match based on the similarity or distance of static data, such as limited information dimensions such as set job requirements, and assess the adaptability of talents and posts through keyword matching, skill label scoring, etc. However, this technical method cannot capture the dynamic changes in the job market (such as regional economic indicator fluctuations, accelerated industry technology upgrading rates, and talent skill proficiency improvements), resulting in a mismatch between the matching results and market demand; for example, a certain skill may be in short supply in the current market, but its future demand may decline sharply in the context of rapid technological iteration; the existing person-post matching technology lacks accuracy and flexibility in predicting future trends, making it difficult for the matching results to meet the actual needs of the dynamic market.

[0004] II. Limitations of existing skill optimization path planning Path planning is one of the important problems in the field of artificial intelligence. In skill optimization path planning, existing methods mostly use greedy algorithms or multi-objective optimization models based on weight adjustment. Greedy algorithms constantly select local optimal solutions (such as selecting the skill or path that improves the fastest at each step), hoping to eventually reach a global optimal or near-global optimal skill optimization solution. This method is often limited to local optimal solutions and cannot globally weigh the difficulty, cost, and potential of skill optimization; for example, a certain skill optimization path may seem efficient in the short term, but may face technical barriers or rapid market demand decline in the long term; in addition, existing talent optimization methods do not include historical inertia resistance, regional restrictions, and other factors in the skill optimization process, resulting in a large deviation between the skill optimization path planning results and actual application scenarios.

[0005] Three, data scarcity in emerging skill areas and insufficient robustness of information entropy diffusion Due to the short development time of emerging skill areas, relevant employment information, skill standards and talent training have not yet formed a scale, resulting in limited number of talents in the market, and the related talent skill information is often incomplete and inaccurate, which makes the data of emerging skill areas scarce; due to the scarcity of data, traditional methods of matching and optimizing talent skills (such as keyword matching and skill labels based on historical data) are no longer applicable in emerging skill areas, resulting in increased difficulty in evaluation; at the same time, it is difficult to accurately understand the development trend and market demand of emerging skill areas, resulting in ambiguity in the direction of talent skill optimization and talent training; more importantly, the existing information entropy perception diffusion process for text generation mostly uses fixed diffusion rate or simple similarity measurement, which makes the information entropy diffusion model unable to effectively handle the problem of associated information completion in data sparse areas; for example, the implicit association between a new skill and existing skills may be ignored due to lack of data, resulting in deviation of matching and planning results; in addition, the existing technology does not fully consider the uncertainty factors in the market environment, such as policy adjustment, economic fluctuations, etc., further limiting the robustness and adaptability of the model. SUMMARY

[0006] To solve the above problems, the present disclosure proposes a talent skill optimization path planning method and system based on dynamic job matching, which greatly improves the accuracy of traditional job matching technology for emerging technology fields and the rationality of emerging technology talent skill optimization path planning method.

[0007] According to some embodiments, the present disclosure adopts the following technical solutions: The talent skill optimization path planning method based on dynamic job matching comprises: Obtaining multi-dimensional data, including skill static state information, market dynamic demand data, job demand information and talent skill score; Performing data physical mapping and introducing inertia index on the multi-dimensional data, constructing an employee skill state space model to represent the potential value, inertia value and potential energy value of each skill; Based on the employee skill state space model, a skill topology graph is constructed with skills as nodes and correlation degree between skills as edge weight, and the energy mapping relationship between multi-skill nodes and neighbor skill nodes in the skill topology graph is integrated to obtain a multi-scale energy network; Through the improved information entropy diffusion model, the skill correlation degree and employee potential value in the multi-scale energy network are completed and optimized to obtain an optimized multi-scale energy network; Using the optimized multi-scale energy network and job demand data, multi-objective particle game optimization is performed to obtain the globally optimal talent skill optimization path.

[0008] According to some embodiments, the present disclosure adopts the technical solutions as follows: The talent skill optimization path planning system based on dynamic human-post matching comprises: The data acquisition module is configured to acquire multi-dimensional data, including skill static state information, market dynamic demand data, post demand information and talent skill score; The state construction module is configured to perform data physical mapping and inertia index introduction on the multi-dimensional data, and construct an employee skill state space model to represent potential values, inertia values and potential energy values of each skill; The network construction module is configured to construct a skill topology graph with skills as nodes and correlation degrees between skills as edge weights based on the employee skill state space model, and integrate energy mapping relationships between multi-skill nodes and neighbor skill nodes in the skill topology graph to obtain a multi-scale energy network; The network optimization module is configured to perform complement optimization on skill correlation degrees and employee potential values in the multi-scale energy network through an improved information entropy diffusion model to obtain an optimized multi-scale energy network; The path optimization module is configured to perform multi-objective particle game optimization on the optimized multi-scale energy network and post demand data to obtain a globally optimal talent skill optimization path.

[0009] According to some embodiments, the present disclosure adopts the technical solutions as follows: A computer program product comprises a computer program, which, when executed by a processor, implements the talent skill optimization path planning method based on dynamic human-post matching.

[0010] According to some embodiments, the present disclosure adopts the technical solutions as follows: A non-transitory computer readable storage medium for storing computer instructions, which, when executed by a processor, implements the talent skill optimization path planning method based on dynamic human-post matching.

[0011] According to some embodiments, the present disclosure adopts the technical solutions as follows: An electronic device comprises a processor, a memory and a computer program; the processor is connected with the memory, and the computer program is stored in the memory; when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the talent skill optimization path planning method based on dynamic human-post matching.

[0012] Compared with the prior art, the present disclosure has the beneficial effects that: 1. By constructing a static energy model based on physical driving (including skill potential energy, inertia and potential optimization energy), and introducing a market demand dynamic energy field and a system resistance model, the present application converts the skill optimization problem into a dynamic energy optimization problem. Unlike traditional static skill tag matching methods, this method can reflect regional economic index fluctuations, industry technology upgrading, and talent skill proficiency, etc. dynamic factors in real time, improving the timeliness and accuracy of the matching results. In a market environment where a certain skill demand is rapidly declining or emerging skill demand is surging, the system can adjust the skill node weight through the dynamic energy field, accurately predict future trends and provide adaptive optimization suggestions.

[0013] 2. The present application models the skill node as a dynamic particle, combines a multi-objective optimization function (including skill optimization cost, income and market adaptability), and performs global path planning based on game optimization algorithm. Compared with traditional greedy algorithm or weight adjustment model, this method can globally weigh the difficulty, cost and potential of skill optimization, avoiding the problem of long-term technical barriers or market demand decay caused by local optimization. At the same time, the model includes historical inertia resistance and regional restriction factors to ensure that the planning results are highly consistent with the actual application scenario. For example, in a certain skill optimization path planning, the system not only considers short-term income, but also evaluates long-term market demand changes through game optimization to provide more forward-looking optimization strategies.

[0014] 3. For the sparsity problem of emerging skill data, the present application introduces an energy diffusion mechanism based on information entropy to dynamically complete and adjust the skill energy distribution. Unlike fixed diffusion rate or simple similarity measurement method, this mechanism can effectively capture the implicit association between emerging skills and existing skills, and complete the associated information in the sparse area through energy diffusion. In addition, the model combines uncertain factors in the market environment (such as policy adjustment and economic fluctuations) to improve the robustness and adaptability of the system. When evaluating the potential value of a certain emerging skill, the system can use the information entropy diffusion mechanism to mine its implicit relationship with existing skills, providing more comprehensive and reliable matching and planning results. BRIEF DESCRIPTION OF DRAWINGS

[0015] The accompanying drawings, which form a part of this disclosure, are used to provide further understanding of the disclosure, and the illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure, and do not constitute improper limitations on the present disclosure.

[0016] Figure 1 is a method flowchart provided by example 1; Figure 2 is a flowchart of constructing an employee skill state space provided by example 1; Figure 3 is a flowchart of constructing a multi-scale energy network provided by example 1; Figure 4is a dynamic diffusion and adjustment flowchart provided by example 1; Figure 5 is a whole actual business processing flowchart provided by example 1. DETAILED DESCRIPTION

[0017] The present disclosure will be further described below in conjunction with the accompanying drawings and examples.

[0018] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present disclosure. Unless otherwise indicated, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present disclosure belongs.

[0019] It should be noted that the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and furthermore, it should be understood that when the terms "comprise" and / or "comprises" are used in the specification, there is a reference to the presence of a feature, step, operation, device, component, and / or combinations thereof.

[0020] Example 1 In one embodiment of the present disclosure, a talent skill optimization path planning method based on dynamic human-post matching is provided, comprising: Step one, obtain multi-dimensional data, including skill static state information, market dynamic demand data, post demand information and talent skill score; Step two, data physical mapping and introduction of inertia index are performed on the multi-dimensional data, and an employee skill state space model is constructed to represent the potential value, inertia value and potential energy value of each skill; Step three, based on the employee skill state space model, a skill topology graph is constructed with skills as nodes and correlation degree between skills as edge weight, and the energy mapping relationship between multi-skill nodes and neighbor skill nodes in the skill topology graph is integrated to obtain a multi-scale energy network; Step four, through the improved information entropy diffusion model, the skill correlation degree and employee potential value in the multi-scale energy network are completed and optimized to obtain an optimized multi-scale energy network; Step five, using the optimized multi-scale energy network and post demand data, multi-objective particle game optimization is performed to obtain a globally optimal talent skill optimization path.

[0021] As an embodiment, the talent skill optimization path planning method based on dynamic human-post matching of the present disclosure greatly improves the accuracy of traditional human-post matching technology for emerging technical fields and the rationality of emerging technical talent skill optimization path planning method, such as Figure 1 As shown in the specific implementation process as follows: S0, acquire multi-dimensional data, including skill static state information, market dynamic demand data, job demand information and talent skill score.

[0022] 1) Skill static state information, represented by skill vector, used to represent the basic hard skills and soft skills required for each skill. The similarity between two skills can be calculated through the skill vector.

[0023] 2) Talent skill score (range 0~100), used to represent the proficiency of employees in each skill.

[0024] 3) Job demand information (range 0~100), used to represent the expected score of the post on the skill.

[0025] 4) Market dynamic demand data: such as HPC demand growth rate, parallel programming market heat, big data derivative application and other related data.

[0026] S1, data physical mapping and inertia index introduction, build employee skill state space model, represent the potential value, inertia value and potential energy value of each skill.

[0027] By extracting the physical connotation of the multi-dimensional talent data, a physical quantity-based employee skill state space is constructed, which provides physical background support for subsequent energy network construction and game decision-making. Therefore, the core of this step is to map traditional numerical features to energy spectrum with physical meaning, and introduce dynamic optimization potential value and inertia index for each skill node, so that traditional static data has the characteristics of time and space dynamic change, such as Figure 2 As shown in the following figure, the specific steps are as follows: S11, data feature extraction and physical connotation mapping First, extract the state features related to skill optimization from multi-dimensional data sources, including regional economic indicators, industry technology upgrading rate, talent skill proficiency, and industry life cycle.

[0028] In the prior art, these data are essentially static numerical indicators. In this embodiment, they are converted into physical characteristics related to energy field through physical mapping.

[0029] The numerical characteristics of each skill node are converted into energy spectrum through a mapping function. Specifically, the skill matching degree is regarded as the potential energy value, which is associated with the supply and demand of the skill and reflects the potential energy change of the skill in the labor market.

[0030] A kind of optimization potential value is proposed, which represents the potential of the skill from the current market state to the target state optimization:

[0031] wherein, is the current matching degree of the skill, represents the dynamic demand of the market, and are the energy values of demand and supply, respectively.

[0032] The current matching degree of the skill here is specifically: For example, the talent skill score of person E1 is data structure = 80, parallel programming = 30, big data = 20, HPC = 0, assuming the demand skill is HPC, according to the similarity of this skill (or other skill indicators) with HPC based on data structure, the cosine similarity is calculated, and the cosine similarity is defined as follows:

[0033] wherein, represents the dot product of vectors, represents the norm of vector S.

[0034] Specifically, based on the mapping function, the energy spectrum gives each skill node a dynamic energy potential, not only considering the current skill matching degree, but also further containing the energy changes that may be encountered in the future optimization process.

[0035] S12, introduce inertia index and momentum characteristics Further, the embodiment proposes an inertia index to simulate the historical accumulation and transition resistance of talents in the skill optimization process.

[0036] This index is very coupled with the historical data of the skill, the learning ability of the talent and the past skill optimization experience, and is very coupled with the inertia concept in physics, which can better represent the resistance in the skill optimization process. Define the inertia value of each talent skill , which represents the momentum resistance that needs to be overcome by the talent in the current skill state optimization:

[0037] wherein, is an adjustment factor, represents the time scale of historical data, represents the skill value of the past time step, similar to the score, for example, the skill score of person E1 is: data structure = 80, parallel programming = 30, big data = 20, HPC = 0, the skill value can be synthesized according to the evaluation results of multiple channels, for example, reference test scores, project results, certificates, peer evaluation or self-evaluation questionnaire, then these evaluations are added or normalized according to certain weights, that is, the quantitative score of each skill at the current time can be obtained, such as "data structure = 80" "parallel programming = 30" such numerical values; Or directly according to the score.

[0038] Based on the inertia value, the time lag effect and optimization resistance in the talent optimization process are simulated, and the momentum characteristics in time and space are injected into the employee skill state space (expanding the perspective dimension) to reflect the dynamic resistance caused by historical factors in the skill optimization process.

[0039] S13. Physical feature fusion and state space construction The energy spectrum and inertia values ​​obtained through the above steps are integrated into a unified physical feature, forming a multidimensional state space for intelligent talent matching and optimization. In this space, each skill node is described not only by physical quantities such as matching degree, optimization potential, and inertial resistance, but also by the dynamic changes in the skill in the labor market.

[0040] When constructing the state space, the physical characteristics of each skill node are integrated to form a unified state representation. , including technical potential energy, kinetic energy, and inertial resistance physical quantities:

[0041] in, is the weight of each physical quantity, It is the state of the external energy field related to skills, reflecting changes in the market environment.

[0042] That is to say, through the fusion of physical features, basic support is provided for the energy network and game decision-making in subsequent steps, ensuring that the status of each skill node can not only reflect the current matching degree, but also reflect the optimization potential and energy changes of skills in a dynamic market environment.

[0043] S14. Output employee skill state space Output employee skill state space It serves as input data for subsequent steps (dynamic skill topology map construction and information entropy diffusion), ensuring that subsequent energy networks and game models can be calculated and optimized based on real and dynamic physical characteristics.

[0044] Current intelligent talent matching and optimization, specifically traditional skill optimization path planning (as reviewed in relevant patent documents and journals), typically relies on static data and quantitative analysis, ignoring market dynamics and historical inertia. This embodiment, however, proposes the concepts of physical state and energy spectrum, imbuing each skill node with dynamic physical properties during the optimization process, more closely resonating with real-world skill optimization characteristics. The proposed physical characteristics of this algorithm can, to a certain extent, address the shortcomings of traditional methods in dealing with dynamic market changes, the complexity, and resistance of skill optimization, providing more accurate optimization path prediction and decision support.

[0045] S2, based on the employee skill state space model, a skill topology graph is constructed with skills as nodes and correlation degree between skills as edge weight, the energy mapping relationship between multi-skill nodes and neighbor skill nodes in the skill topology graph is integrated, and a multi-scale energy network is obtained.

[0046] The objective of this step is to formulate the talent intelligence matching and optimization physical characteristics based on the output of step S1, a dynamic skill topology graph is constructed by proposing an improved network architecture, and a multi-scale energy network is formed. The network can not only represent the relationship between skills, but also reflect the dynamic changes caused by energy gradient, inertial resistance and technical activation energy in the skill optimization process.

[0047] That is, through this construction method, the energy difference and barrier between emerging skills and traditional skills can be more realistically described, providing a physical background for subsequent information entropy diffusion and game decision.

[0048] Figure 3 is a dynamic skill topology architecture diagram of a dynamic post matching and new talent skill optimization method based on multi-objective game and information entropy diffusion model provided by the embodiment, as Figure 3 The dynamic skill topology architecture of the dynamic post matching and new talent skill optimization method based on multi-objective game and information entropy diffusion model in the embodiment can include the following contents.

[0049] In some optional embodiments, the S2 step can be specified as S21-23, as Figure 3 shown, the specific contents are as follows: S21, construction of talent-market skill topology graph Based on the physical characteristics (energy potential, inertia value) of each skill node obtained in step S1, the skill topology graph is constructed, which is actually to calculate the multi-dimensional feature information X node of each skill node.

[0050] That is, in this topology graph, each node not only represents a skill or post, but also becomes an entity carrying physical attributes ("energy state", "inertia value" and "local diffusion coefficient").

[0051] That is, each skill node carries dynamic physical characteristics, including optimization potential , inertia value , and local diffusion coefficient related to skills .

[0052] Through the fusion of physical characteristics, a comprehensive physical attribute representation is constructed for each node:

[0053] wherein, is the weighted coefficient of physical characteristics, is the local diffusion coefficient calculated according to the optimized ability and resistance of the skill around the node, reflecting the degree of difficulty of the skill node in the network diffusion.

[0054] The embodiment is intended to form a dynamic skill topology map by quantifying and integrating the physical characteristics of each node, so that the network structure not only reflects the relationship between skills, but also truly reflects the physical changes experienced in the optimization process, that is, the degree of difficulty.

[0055] S22, construction of edges between nodes and calculation of energy gradient Calculate the edge weight W between skill nodes ij Organize the nodes into a skill topology map, and the topology map only has the relationship between skills and skills, that is, the edge relationship between nodes is constructed, and the edge weight is no longer dependent on the traditional similarity index, and is further dynamically adjusted based on the energy gradient scale. The edge weight between nodes is determined by the local skill optimization potential, skill dissipation barrier, and activation energy of regional technology upgrading.

[0056] For any two skill nodes and , the edge weight is calculated by the following formula:

[0057] wherein, is the difference in skill optimization potential between nodes, is the energy consumption caused by skill dissipation, optimization barrier and other factors, is an adjustment factor for balancing the degree of difficulty of local skill optimization.

[0058] which can be further expressed as:

[0059] wherein, is an adjustment factor that controls the weighting degree of energy difference between skill nodes, and are the optimization potentials of nodes and , respectively.

[0060] Vᵢ(t) and Vⱼ(t) are actually Vtransfer-j(t) and Vtransfer-i(t), which are the optimization potentials calculated in step S11, is the difference between the optimization potential values of the two nodes.

[0061] Optimization potential value: that is, the energy potential difference contained when the skill node jumps from the current state to the target state, the higher the value, the greater the optimization space of the node.

[0062] The edge weight between nodes can dynamically reflect the energy difference and optimization barrier between skills, and can more accurately capture the "barrier" relationship between skills through the calculation of energy gradient and resistance in this embodiment.

[0063] S23, construction of a multi-scale energy network The energy state of the node and the adjacent node is calculated to depict the coupling relationship between the multi-skill node and the adjacent skill node. The rule of the adjacent node comes from the edge calculation in S22, that is, the edge weight between the two skill nodes is not zero, that is, they are adjacent to each other. Based on the construction results of the above nodes and edges, a multi-scale energy network is further constructed, which can reflect the complex energy transmission and energy difference between skills.

[0064] Specifically, in the multi-scale energy network, the energy state and energy gradient information of the node are transmitted to the adjacent node, and are adjusted according to the edge weight between the nodes and the energy gradient. The energy state of the node is dynamically evolved through the following equation:

[0065] Wherein, in order to calculate an energy value obtained by the node at moment, is the edge weight between the node and its adjacent node Each skill node has an initial energy state E i (0) and E j (0), the initial energy is a predetermined value, which can be calculated by the sum of the scores of the skills, is the maximum energy value that can be reached in the network, generally a threshold value is given.

[0066] The energy growth of node i mainly comes from the energy "radiation" of adjacent node j, and the growth amplitude is gradually limited by the upper limit of the node itself. When node i gradually approaches , becomes very small, and the energy value of node i also tends to be saturated and no longer increases greatly; then gradually iterate, and the iteration process is the dynamic evolution of .

[0067] That is, based on the multi-scale energy network, the energy distribution and optimization potential of the skill node are dynamically adjusted through the energy exchange and diffusion mechanism between nodes, and physical field support is provided for the subsequent information entropy diffusion module and game decision.

[0068] Through the above steps, the multi-scale energy network is finally output As the base model in the subsequent step. The model can accurately represent the energy difference between skills and positions, the barrier relationship and the local energy field.

[0069] Traditional skill network construction is usually based on static similarity or distance measurement, which cannot dynamically reflect the energy difference and resistance between skills. However, the technical solution proposed in this embodiment can not only reflect the energy changes in real time during the skill optimization process, but also has spatiotemporal dynamics. Compared with traditional methods, the embodiment can more realistically depict the optimization barriers and energy differences between emerging skills and traditional skills.

[0070] S3, through the improved information entropy diffusion model, the skill correlation and employee potential value in the multi-scale energy network are completed and optimized, and the optimized multi-scale energy network is obtained.

[0071] The goal of this step is to improve the existing information entropy diffusion model based on the multi-scale energy network constructed in step S2, introduce energy driving mechanism, local adaptive diffusion rate adjustment and local oscillation monitoring, optimize skill correlation and data scarcity in the diffusion process, and obtain the multi-scale energy network after energy correction.

[0072] S31 and S32 perform information entropy diffusion through energy driving and local adaptive methods, and S33 outputs the optimized potential evaluation data of each skill node after energy correction. Essentially, it is an iterative correction of the node energy distribution in S23, outputting new Enode(t) values.

[0073] The information entropy diffusion module proposed in this draft not only enhances the physical driving synchronization step S1S2 content on the basis of traditional entropy diffusion, but also provides robustness and flexible response ability through local energy gradient adjustment, and finally outputs skill correlation and optimization potential evaluation data. Combined with historical data, the sparse data area is completed. For example, the vacancy degree of the HPC skill node is diffused to the parallel programming or big data nodes associated with it, so as to supplement the data sparse area (such as HPC).

[0074] The scarce data to be supplemented or made up is the data that cannot be fully obtained or is not complete in the early stage (S1, S2), but affects the accuracy of skill optimization and matching, including (but not limited to) the following categories: 1. Historical use / learning data of emerging or unpopular skills Emerging skills often lack mature learning curves, market feedback or historical data, making it impossible to have sufficient quantitative references like traditional skills.

[0075] 2. Local missing information in talent portrait After employees change positions or jobs, there is a gap in their skill development or training experience; or external talents do not have complete historical ability assessment records.

[0076] 3. Incomplete data at the industry or regional level There are few statistics on emerging industries, technology growth rates and regional economic indicators, the statistical caliber is not unified, and the time span is insufficient, making it difficult to support accurate modeling.

[0077] 4. Lack of skill relevance / synergy in sparse scenarios When there are few overlapping scenarios between skills and insufficient input data is still in the initial stage, it is impossible to directly give a reliable similarity or synergy coefficient.

[0078] 5. Micro-data such as market demand gap and salary distribution New positions and new occupations have not yet accumulated sufficient recruitment demand, salary distribution or job competency indicators, resulting in data gaps.

[0079] The energy-driven and locally adaptive information entropy diffusion mechanism is introduced here to reasonably supplement and infer the missing or insufficient data through the energy gradient, inertia index, diffusion resistance and other physical information of the surrounding nodes (skills / talents / positions), so as to improve the robustness and accuracy of emerging skills and sparse data areas.

[0080] In some optional embodiments, step S3 may be described in detail as S31-34, with the specific contents as follows: S31, information entropy, energy density coupling mechanism First, the information entropy diffusion model is improved so that it is not only a dimensionless indicator, but is coupled with the "energy density" of the node and the local "diffusion damping" to achieve more accurate modeling of the skill optimization process.

[0081] In a multi-scale energy network, let the entropy of each skill node be Adjusted by the energy density and diffusion damping coefficient. Specifically, the diffusion rate of information entropy is determined by the local energy density of the node and damping coefficient Decide:

[0082] in, is the adjustment factor, For nodes The energy density, is the local diffusion damping coefficient of the node, is the spatial diffusion operator of entropy.

[0083] is based on It is determined by the physical characteristics of the node itself, such as momentum, inertia, energy potential, etc. (i.e., the characteristic parameters inherited from S1 and S2), and is initially a preset value.

[0084] This mechanism couples the energy density of nodes with the diffusion damping coefficient to ensure that the diffusion process of information entropy in the skill network not only takes into account the energy distribution of nodes, but also dynamically adjusts the diffusion rate, thereby improving the robustness of data sparse areas.

[0085] S32, local adaptive diffusion rate adjustment In this embodiment of the energy network, the node diffusion rate is not fixed but dynamically adjusted based on the local energy gradient. This ensures that in areas with sparse skill data, missing data can be supplemented through the guidance of the energy gradient, while avoiding the uncertainty caused by excessive diffusion.

[0086] Diffusion rate of each node Based on the local energy gradient Automatic adjustment. Specifically, the diffusion rate is adaptively updated using the following formula:

[0087] in, and is the adjustment factor, is a node The energy gradient around It is the second-order derivative of local energy and reflects the rate of change of the energy field. The energy gradient refers to the direction and magnitude of the energy change in the node neighborhood, that is, the first-order spatial derivative of the energy distribution.

[0088] Through this adaptive diffusion rate adjustment mechanism, the entropy diffusion rate is dynamically adjusted according to the changes in the local energy gradient, thereby ensuring that missing information can be more effectively supplemented in areas where skill data is scarce, while ensuring the balance of global energy distribution.

[0089] S33. Local oscillation monitoring and energy balance adjustment Based on the above sub-steps, to further enhance the model's adaptability during skill optimization, this step proposes a local oscillation feedback monitoring mechanism to detect physical resonance or energy imbalances that may occur during skill optimization. If energy oscillation or imbalance is detected in a local area, this embodiment automatically introduces an assisted diffusion sub-module to adjust the local energy balance.

[0090] First, a local oscillation factor is proposed , representing a node Is there any energy imbalance or oscillation in the surroundings? The factor is calculated using the following formula:

[0091] wherein, is the local damping coefficient, reflecting the energy consumption degree of the node. If exceeds the set threshold , the local energy balance adjustment mechanism is started.

[0092] For the nodes that appear oscillation, an auxiliary diffusion sub-module is proposed to adjust the local energy state, so that it tends to balance. The auxiliary diffusion module update formula is:

[0093] wherein, is the adjustment factor.

[0094] The updated Enode(t), ∇²Enode(t) and (t) exceed the threshold, the local energy balance adjustment is started, and the updated value is output. The auxiliary diffusion sub-module is specially designed for these nodes that exceed the threshold to correct the energy field, so that it returns to balance.

[0095] Energy driving is the process of promoting the continuous evolution of the node's own energy by the energy difference brought by the neighbor nodes and network structure, which can be directly understood as the difference of the absolute value of the node.

[0096] The introduction of local oscillation monitoring and energy balance adjustment can automatically adjust the energy distribution in the network when energy imbalance occurs in the skill optimization process, thereby avoiding abnormal behavior in the information entropy diffusion process and ensuring the stability and robustness of the network.

[0097] S34, output skill association and optimization potential evaluation data Through the above steps, the information entropy diffusion module proposed in this embodiment can output skill association and optimization potential evaluation data after energy correction and scene mapping. These data provide data-driven support for subsequent game decision and skill optimization path planning.

[0098] The data output in this step not only contains the association degree between skills, but also contains the potential evaluation of skill optimization, which can provide accurate guidance for talent skill optimization and job matching.

[0099] Figure 4 used to show the dynamic diffusion and adjustment process of combining information entropy and energy field. The input part comes from the topological network parameters constructed in the previous steps, including the energy gradient between skill nodes, the inertia value and the local diffusion coefficient. In the "entropy-energy" coupling modeling stage, the index E·lnD is integrated to include the energy density E and the diffusion rate D of the node into the same model, realizing the physical description of information entropy.

[0100] The coupling process generates feedback to dynamically adjust the diffusion speed or strategy for subsequent signal transmission to the "local adaptive diffusion" module, so as to keep the balance between information entropy diffusion and energy distribution at all times.

[0101] On the basis of coupling modeling, the diffusion behavior between nodes is further processed in detail through local adaptive diffusion and gradient acceleration rules. If the energy gradient of some region is large, the diffusion process will be appropriately accelerated, and vice versa, so as to ensure that the necessary supplement can be obtained in the places where data is sparse or skills are missing.

[0102] Once the oscillation monitoring detects interference or frequent fluctuations in energy distribution, it sends an abnormal signal to the sparse damping adjustment to trigger the inhibition or correction of local diffusion, preventing the energy from fluctuating greatly and affecting the accuracy of entropy diffusion.

[0103] Finally, the feedback and interaction between the links will make the coupling process of energy and entropy tend to be stable, and the output end will give the entropy distribution after the energy field is corrected.

[0104] The traditional information entropy diffusion method usually relies on static entropy values and simple similarity measures, and cannot fully consider the energy differences and dynamic changes between skills. The technical solution proposes an energy-driven mechanism, local adaptive diffusion rate adjustment, and local oscillation monitoring, so that the information entropy diffusion process is more accurate, dynamic, and can adaptively cope with data sparsity and energy imbalance.

[0105] In the business background of talent intelligent matching and optimization, the present solution proposes a method based on energy network, which can better simulate the potential energy barriers and resistance in the skill optimization process.

[0106] S4, using the optimized multi-scale energy network and job demand data, multi-objective particle game optimization is performed to obtain the globally optimal talent skill optimization path.

[0107] Based on the concepts of "energy cost" and "barrier crossing", the skill nodes are regarded as particles moving in an energy field, the energy barriers that need to be overcome for the skill to jump from the current potential energy to the target potential energy region are weighed, the difficulty and energy consumption of the skill optimization path are quantified, and the globally optimal talent skill optimization path is output; therefore, the goal of this step is to build an improved multi-objective game model based on the skill optimization data output in S3 after physical field correction. Unlike the traditional game model which adjusts the target weight alone, this step cooperates with the above-mentioned energy cost and barrier crossing concepts, regards each talent and skill node as a particle moving in an energy field, and weighs the energy barriers that need to be overcome for the skill to jump from the current potential energy to the target potential energy region.

[0108] The system dynamically evaluates the energy consumption of the skill optimization path by simulating the particle motion and based on the idea of energy minimization, so as to find the optimal skill optimization path and effectively avoid the local optimization problem, and realize the global most stable dynamic balance.

[0109] In some optional embodiments, the S4 step can be specified as S41-44, and the specific content is as follows: S41, define energy cost and barrier crossing Further, before constructing the game decision model, the embodiment proposes an "energy cost" and "barrier crossing" algorithm physical feature, which is actually to quantify the optimization difficulty and energy consumption of each skill optimization path, and provides a basis for the subsequent decision process.

[0110] Energy cost : represents the energy consumed in the optimization process from the skill node to the target node . Its definition is the energy barrier that needs to be overcome by the energy gradient adjustment of the node to reach the target node :

[0111] Wherein, is the energy gradient of the node, reflecting the change of the energy field; is the speed of particle motion, indicating the transition rate of the skill node in the energy field.

[0112] Barrier crossing : represents the energy barrier that needs to be overcome by the skill node when transitioning to the target node . The barrier is determined by the "potential energy" difference of the target node and the physical resistance in the optimization process:

[0113] Wherein, and are the potential energy of the target skill node and the source skill node respectively, is the adjustment factor, is the inertia value of the node, reflecting the physical resistance in the optimization process.

[0114] That is, the energy cost and barrier crossing proposed in the embodiment provide a quantitative energy consumption model for the subsequent game decision, helping the system to evaluate the actual difficulty and required resources of each skill optimization path.

[0115] S42, construction of multi-objective game model After the energy cost and barrier crossing are determined, a multi-objective game model is further constructed. Unlike the traditional game model which adjusts weights, the present scheme integrates multiple decision indicators (including matching degree, optimization cost, and skill improvement potential) into a unified energy function, which represents system energy consumption and dynamic balance. Based on this function algorithm, the system can guide the selection of the optimal path during the simulation of particle motion.

[0116] Specifically, the objectives of each skill optimization path are integrated into a unified energy function , which includes multiple target decision indicators:

[0117] wherein, is the energy consumption, is the barrier crossing, is the skill improvement potential, is the corresponding weight coefficient.

[0118] Based on the integration of multiple target decision indicators, the energy consumption, difficulty, and potential of each skill optimization path can be more comprehensively evaluated, ensuring the selection of the path with the most dynamic balance.

[0119] S43, particle motion simulation and path optimization By simulating the motion of particles in an energy field, the energy consumption of each skill optimization path can be dynamically evaluated, and the optimal optimization channel can be found through the principle of energy minimization. Particle motion simulation can avoid local optimal solutions and ensure the global optimality of path selection. That is, each skill node Like "particles" moving in an energy field, the optimization process is simulated by the following particle motion equation:

[0120] wherein, is the position of the particle, representing the state of the skill node, is the velocity of the particle, representing the optimization rate in the skill optimization process.

[0121] The motion of particles in the energy field is influenced by the energy gradient, energy consumption, and barrier crossing. Ultimately, the optimal skill optimization path is selected by minimizing the total energy consumption.

[0122] By simulating the dynamic motion of particles in an energy field, the system can automatically optimize the skill optimization path and find a skill optimization scheme with global optimality, avoiding the local optimal problems that may occur in traditional methods.

[0123] S44, output optimization strategy and evaluation data Finally, the system outputs an optimized strategy scheme based on the physical energy model, each path not only has a numerical ranking, but also has detailed potential energy state and energy cost physical quantity, to truly reflect the dynamic resistance and incentives faced during actual optimization.

[0124] In fact, the traditional game model usually only adjusts the target weight to solve the optimal path, which cannot effectively consider the energy barriers and dynamic changes in the optimization process. The technical solution can find the global optimal solution in a dynamic environment through the energy cost and barrier crossing idea and the simulation of particle motion, avoiding the problem of local optimization. Given the business background of talent intelligent matching and optimization, the technical solution can accurately evaluate the skill optimization path according to the physical characteristics (such as energy cost, optimization potential, etc.) of each skill node.

[0125] S5, bidirectional feedback and closed-loop energy linkage The goal of this step is to build a bidirectional energy transmission system with a physical field feedback mechanism, link the information entropy diffusion module and the energy game module, and realize a bidirectional closed-loop adjustment to output an iterative correction and dynamic balance talent skill optimization path to adapt to the changing market environment and data distribution.

[0126] That is, unlike the traditional optimization method that relies on weight feedback, the technical solution uses energy feedback and physical field adjustment to enable the optimization network to achieve optimal balance through dynamic adjustment of physical state and energy field when facing market, data uncertainty, and optimization evolution, rather than relying solely on single target weight adjustment.

[0127] In some optional embodiments, the S5 step can be detailed as S51-54, with the specific content as follows: S51, construction of physical field feedback mechanism This step first constructs a physical field feedback mechanism to enable the information entropy diffusion module and the energy game module to conduct bidirectional energy transmission, thereby ensuring that the system can dynamically adjust its response to adapt to the changing market environment and data distribution. Two key physical quantities in the feedback mechanism are energy cost feedback and local energy balance feedback , which come from the game decision module and the diffusion module, respectively. Based on this feedback information, the parameters inside each module can be adjusted, and the dynamic optimization of the system can be driven. Energy cost feedback reflects the energy consumption required for the skill optimization path calculated by the game decision module, which is mathematically represented as:

[0128] wherein, represents the energy gradient of the node, Optimize the rate of skill nodes in the energy field.

[0129] Local energy balance feedback The output from the diffusion module represents the fluctuations in the local energy of the node, which is used to adjust the physical constraints in the energy game model:

[0130] where, represents the second derivative of the energy field, reflecting the fluctuations and imbalances in local energy.

[0131] The feedback mechanism of energy cost and local energy balance is designed to ensure that the information entropy diffusion module and the energy game module can maintain system dynamic balance when facing data uncertainty and optimization changes.

[0132] S52, two-way feedback regulation mechanism After establishing the physical field feedback mechanism, a two-way feedback regulation mechanism is constructed to ensure that feedback information can be transmitted between the two modules.

[0133] When the game module determines a skill optimization path, the energy cost and local energy balance state evaluated by it is fed back to the diffusion module, prompting the diffusion module to adjust the local diffusion coefficient and energy gradient according to the feedback information:

[0134] where, is the local diffusion rate, and is the adjustment factor, which controls the degree of influence of feedback information in the diffusion module.

[0135] At the same time, the energy gradient and local fluctuation information output by the diffusion module are used as additional constraint terms for the game decision module, updating the target energy function, so that the game module can re-evaluate according to the new physical state. The updated energy function is:

[0136] where, is the adjustment factor, representing the influence of local energy fluctuation information on the energy function.

[0137] The two-way feedback regulation mechanism is designed to enable the two modules to interact and adjust in real time, achieving energy conduction and dynamic balance within the system, thereby improving the system's response to market and data changes and ensuring the accuracy and adaptability of the final decision.

[0138] S53, closed-loop energy linkage optimization On the basis of the two-way feedback mechanism, this step further forms a closed-loop energy linkage optimization system. Through the feedback cycle of physical field and energy game, the system can realize adaptive adjustment and optimization when facing data uncertainty and dynamic changes in skill optimization. That is, the closed-loop optimization core is to continuously adjust the parameters in the information entropy diffusion module and the energy game module through the feedback of energy cost, barrier crossing, and local equilibrium state, so that the system can respond to market changes in real time, and ultimately achieve a globally optimal balance state:

[0139] wherein, is the change of the energy state of the system, is the change of the energy state of a single node, is the change of the local energy equilibrium state.

[0140] Through closed-loop energy linkage optimization, the system can automatically adjust the physical field and energy function when facing external uncertainty and optimization evolution, and ultimately maintain an optimal dynamic balance state, avoiding the local optimal trap in traditional methods.

[0141] S54, output optimization decision scheme Finally, through the closed-loop energy linkage system, an optimization decision scheme based on physical field feedback and game decision is output.

[0142] Each path not only has a numerical ranking, but also has detailed "energy cost" and "local energy equilibrium state" physical quantities, which truly reflect energy consumption, dynamic resistance and incentives in the optimization process.

[0143] Based on this embodiment, an efficient skill optimization decision scheme can be provided to help decision makers make the best skill optimization path selection in a dynamic environment and improve the efficiency of talent skill matching and job optimization.

[0144] Traditional optimization methods rely on static weight adjustment and cannot fully consider the dynamic changes of the physical field and data uncertainty. The technical solution proposes a two-way feedback mechanism and a closed-loop energy linkage optimization system, so that the two modules can be adjusted in real time, thereby realizing the dynamic balance and optimization of the system. It can effectively respond to market changes, the dynamic nature of skill optimization, and data sparsity, and provide more accurate and optimal decision support for skill optimization path selection in practical applications.

[0145] Figure 5 is the logical diagram of the business processing of the embodiment, showing the overall actual business processing flow.

[0146] From the basic information submitted by the user, the overall logic of combining the employee skill state space described by physical quantities with subsequent multi-scale energy network and game decision model is embodied.

[0147] When the user uses this embodiment, after inputting the current position stage, work experience and target industry demand, based on the method proposed in S1, the multi-dimensional data is physically mapped and feature extracted, the skill matching degree of each talent, market demand and other static numerical values are converted into "energy spectrum" and "inertia value" which can dynamically evolve in space and time, so as to form an initial state space model of "employee-skill".

[0148] After obtaining the skill state data with energy potential and historical inertia, according to the steps described in S2, a dynamic skill topology graph is constructed and a multi-scale energy network is formed. Through the calculation of energy gradient and local diffusion coefficient, the model can not only capture the potential skill supply and demand difference, but also outline the energy barrier faced by talents when migrating to target posts or upgrading skills.

[0149] Next, combined with the improved entropy diffusion module proposed in S3, the incomplete or sparse data basis is adaptively completed by energy driving, and possible energy oscillation or imbalance phenomenon is monitored to ensure that each skill node is more accurately matched and potential evaluated.

[0150] After summarizing the energy and entropy distribution information, enter the multi-objective game stage of S4. At this time, the user's target conditions and industry indicators are combined to measure the energy cost and barrier crossing of skill transition, and based on the energy minimization idea, the feasible career optimization channel is determined. Finally, under the closed-loop linkage mechanism of S5, the latest game decision and energy distribution are continuously adjusted in both directions, and a strategy scheme with quantitative energy indicators is output.

[0151] Finally, a multi-dimensional interpretation of short-term growth, long-term potential and external environmental changes is output, so as to help the user effectively plan skill improvement and post migration in a dynamic market environment.

[0152] Embodiment 2 In order to better understand the technical solutions of the present application, a specific example is provided, and the numerical value is only for demonstration as an example: A high-tech enterprise found that high-performance computing (HPC) has become an emerging and important skill, but the related experience in the domestic and foreign talent market is limited, and the training path is not clear. The enterprise hopes to help existing employees upgrade to the HPC technology level quickly with the help of this scheme, and at the same time, consider the individual training cost and business demand in the process.

[0153] The company currently has 5 employees (E1~E5), each with different existing skills (such as data structures, parallel programming, big data platforms, etc.), but little direct experience in HPC skills. The company has 3 positions (R1, R2, R3) with varying degrees of demand for HPC.

[0154] The detailed implementation logic and data flow of this embodiment are as follows: S1: Data feature extraction and physical mapping (1) Input data 1) Skill static state information, represented by a skill vector, used to represent the basic hard skills and soft skills that need to be mastered, for example, skill The skill static state information of , where the basic hard skills and soft skills include: CPU / GPU architecture of computer architecture, memory hierarchy (cache consistency), parallel computing model (SIMD / MIMD); Linux system management, process / thread scheduling, file system, I / O optimization of operating system; C / C++ (mainstream HPC language), Fortran (traditional scientific computing), Python (auxiliary script) of programming language; Matrix operation, graph algorithm, sparse data structure, time complexity optimization of algorithm and data structure; Message passing interface (point-to-point communication, collective communication) of MP, topology mapping; Shared memory parallelization (instruction set, task scheduling) of OpenMP, thread synchronization; CUDA / HIP GPU programming model (Kernel design, memory hierarchy), performance optimization (merge access, BankConflict); Performance analysis tools: Intel VTune, NVIDIA Nsight, Gprof (hot spot location, bottleneck analysis).

[0155] 2) Employee existing skill proficiency (range 0~100) - E1: data structure = 80, parallel programming = 30, big data = 20, HPC = 0 - E2: data structure = 60, parallel programming = 75, big data = 40, HPC = 0 - E3: data structure = 50, parallel programming = 50, big data = 60, HPC = 10 - E4: data structure = 70, parallel programming = 20, big data = 45, HPC = 0 - E5: data structure = 40, parallel programming = 65, big data = 30, HPC = 5 3) Job requirements (R1, R2, R3) expectations for HPC skills - R1: HPC proficiency expectation > 50 - R2: HPC proficiency expectation > 70 - R3: HPC proficiency expectation > 30 4) Industry development data: such as HPC demand growth rate, parallel programming market heat, big data derivative applications, and other related data.

[0156] (2) Processing and output 1) Cleaning and standardization: mapping employee skills, job requirements, etc. to a unified scale; 2) Physical energy spectrum mapping: multiply each skill proficiency by the industry demand growth rate to get potential energy. For example, E2's parallel programming originally scored 75, combined with higher market demand (such as a coefficient of 1.2), it maps to a "potential" of 90; 3) Calculate "inertia value": define resistance based on employee's past learning experience or job transfer. For example, E1 has rich experience in cross-department, with a low inertia value (0.7); E4 has a general acceptance of new technology, with a high inertia value (0.9); 4) Output: form a "employee skill state space", where each employee has a "potential energy" and "inertia coefficient" physical description at each skill node.

[0157] S2: Dynamic skill topology graph and multi-scale energy network construction (1) Input data 1) Energy and inertia information corresponding to each skill node (data structure, parallel programming, big data, HPC) output by S1; 2) Internal or external assessment of the degree of association between skills, such as parallel programming and HPC have a high degree of association in theory and practice, and big data and parallel programming also have a moderate degree of association.

[0158] (2) Processing and output 1) Construct skill nodes: treat "data structure", "parallel programming", "big data", and "HPC" as four core nodes, and record their industry demand and potential value to individual employees; 2) Edge weight: parallel programming and HPC have the strongest association, set a higher edge weight (such as 0.8), meaning it is easier to transfer to HPC if you master parallel programming; - The association of Big Data and HPC can be 0.5, as there are more application-level couplings between the two; - The association of data structure and HPC is relatively weak, which can be set to 0.3; 3) Multi-scale energy network: integrate the proficiency of employees on each node (energy value) and edge relationship, form a topology that can reflect the difficulty of transformation; 4) Output: a visualization skill map: nodes represent skills, edges represent transferable efficiency; employee mapping: E1 has a potential of 80 and inertia of 0.7 on the data structure node, but has no direct mastery of the HPC node, but there is a potential bridge with parallel programming.

[0159] S3: Improve information entropy diffusion (energy-driven and adaptive) (1) Input data 1) Multi-scale energy network output by S2; 2) Incomplete data of employees on certain skills, such as E3 self-studied HPC introduction, actual proficiency is not clear, which needs to be filled or corrected by the model; 3) Industry trends: such as the additional demand for HPC in cloud computing rising by 5% in the past month.

[0160] (2) Processing process and output 1) Information entropy diffusion: diffuse the vacancy of HPC skill nodes to the parallel programming or big data nodes associated with them, thereby supplementing in the sparse area (such as HPC). If E3 has high scores on parallel programming and big data, it is inferred that his HPC starting point may be higher than 0 (the model automatically estimates it to be 15-20); 2) Local adaptation: find that HPC demand has increased sharply, so the diffusion rate of related nodes is accelerated to help the network focus on supplementing and correcting the potential value of HPC; 3) Output: updated skill potential and association, such as E3's HPC score from 10 to 18, and E5 from 5 to 8; A tip report points out that HPC is a "new high-potential skill" and the company should prioritize employee parallel programming -> HPC or big data -> HPC training channels.

[0161] S4: Multi-objective game strategy and optimization path of new skills (1) Input data 1) The latest assessment of employees' HPC skills and proficiency in other skills output by S3; 2) Position R1, R2, R3's different level of demand for HPC; 3) Enterprise training resource constraints: such as 40 hours of HPC training courses per month; employee tolerance for training time and overtime; external instructor budget, etc.

[0162] (2) Process and output 1) Multi-objective game: consider "training cost (time and budget)" "job matching degree" "future transformation benefit" and other factors to find the optimal HPC skill improvement path; 2) Example: if E2 turns to HPC, it needs to strengthen parallel programming from 75 points to 90, and then connect HPC special courses, which is expected to make HPC reach 60 points within 2 months; if E4 turns to HPC, since parallel programming is only 20 points, it needs more time (such as 3 months) to improve HPC to 50 points; 3) The system compares and scores the possible "skill upgrade path" of each person, and finally outputs a "globally optimal" person-job matching solution.

[0163] For example, the most suitable candidate for R2 (HPC ≥ 70 points) is E2; R1 (HPC ≥ 50 points) can be taken by E3; E4 may not be suitable for HPC high-level positions temporarily, but can remain in big data related positions; Output: Clear HPC training path recommendation: who learns parallel programming first, who learns HPC directly, who participates in cloud computing projects at the same time; target time length for each employee to invest, expected HPC score to achieve, and additional resources such as external instructors.

[0164] S5: Two-way feedback and closed-loop energy linkage (focus on new skill iteration) (1) Input data 1) The optimal "HPC skill improvement" plan given by S4; 2) Training feedback during actual execution: if E2's HPC score is only improved from 0 to 30 after 2 weeks of learning (lower than the expected 35), or E3's training is delayed due to busier projects; 3) New HPC variant demands in the market (such as HPC+AI modules) may bring new skill initial scores and relatedness.

[0165] (2) Process and output 1) According to the training progress and demand changes, the information entropy diffusion module will update the potential value or resistance coefficient of HPC and its related nodes (parallel programming, big data), so that the network is more in line with the latest situation; 2) Multi-objective game re-evaluation: if E2 progress is hindered, the system may consider optimizing the training program or assigning different training content for E2 and E1; 3) Output: New round of "skill optimization strategy": dynamically supplement or resource tilt for HPC training duration; Remind management of new HPC+AI needs, and reserve learning in this direction for employees with high parallel programming scores (E2 or E5) in advance if necessary; Build a continuous iteration "closed loop system" to allow the enterprise to fine-tune according to external market and internal progress when introducing new skills (such as HPC).

[0166] The enterprise regards HPC as a new skill and focuses on how to enable existing employees to quickly and effectively master HPC and meet job requirements at each step; collect and physically map employee skill data to form a basic "energy and inertia" description; build a skill topology network with "energy gradient" and "conversion difficulty" attributes to provide intuitive transferable relationships for HPC; improved information entropy diffusion compensates for the sparsity of HPC data and dynamically adjusts the potential "energy value" of HPC according to market demand; balance training costs, business needs, and employee personal factors in multi-objective games to provide HPC path optimization solutions; through actual training feedback and market changes, update the above network and game strategy in a closed loop to make HPC talent training adaptive and sustainable.

[0167] Embodiment 3 A talent skill optimization path planning system based on dynamic human-post matching is provided in an embodiment of the present disclosure, comprising: A data acquisition module configured to acquire multi-dimensional data including skill static state information, market dynamic demand data, job requirement information, and talent skill score; A state construction module configured to perform data physical mapping and inertia index introduction on the multi-dimensional data, and construct an employee skill state space model to represent potential value, inertia value, and potential energy value of each skill; A network construction module configured to construct a skill topology graph with skills as nodes and correlation degree between skills as edge weight based on the employee skill state space model, and integrate energy mapping relationships between multi-skill nodes and neighbor skill nodes in the skill topology graph to obtain a multi-scale energy network; A network optimization module configured to complete and optimize the skill correlation degree and employee potential value in the multi-scale energy network through an improved information entropy diffusion model to obtain an optimized multi-scale energy network; The path optimization module is configured to perform multi-objective particle game optimization by using the optimized multi-scale energy network and the post demand data, and obtain a globally optimal talent skill optimization path.

[0168] Embodiment 4 In an embodiment of the present disclosure, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the talent skill optimization path planning method based on dynamic human-post matching.

[0169] Embodiment 5 In an embodiment of the present disclosure, a non-transitory computer-readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the talent skill optimization path planning method based on dynamic human-post matching.

[0170] Embodiment 6 In an embodiment of the present disclosure, an electronic device is provided, comprising a processor, a memory, and a computer program; wherein the processor is connected with the memory, and the computer program is stored in the memory; when the electronic device is running, the processor executes the computer program stored in the memory, so that the electronic device implements the talent skill optimization path planning method based on dynamic human-post matching.

[0171] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks. Figure 1 The functions specified in one or more flows and / or blocks.

[0172] These computer program instructions can also be loaded onto a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable data processing device to produce a computer-implemented process, so that the instructions executed by the computer or other programmable data processing device provide a process for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks. Figure 1 The functions specified in one or more flows and / or blocks.

[0173] The specific embodiments of the present disclosure are described above with reference to the accompanying drawings, but are not intended to limit the protection scope of the present disclosure, and those skilled in the art should understand that various modifications or changes made on the basis of the technical solutions of the present disclosure without creative labor are still within the protection scope of the present disclosure.

Claims

1. A method for planning a talent skill optimization path based on dynamic person-post matching, characterized in that, The method comprises the following steps: acquiring multi-dimensional data, including skill static state information, market dynamic demand data, post demand information and talent skill score; performing data physical mapping and introducing inertia index on the multi-dimensional data, and constructing an employee skill state space model to represent potential value, inertia value and potential energy value of each skill; based on the employee skill state space model, constructing a skill topology graph with skills as nodes and correlation degree between skills as edge weight, integrating energy mapping relationship between multi-skill nodes and neighbor skill nodes in the skill topology graph, and obtaining a multi-scale energy network; completing and optimizing the skill correlation degree and employee potential value in the multi-scale energy network through an improved information entropy diffusion model, and obtaining an optimized multi-scale energy network; using the optimized multi-scale energy network and post demand data to perform multi-objective particle game optimization, and obtaining a globally optimal talent skill optimization path.

2. The talent skill optimization path planning method based on dynamic person-post matching according to claim 1, wherein, The data physical mapping and inertia index introduction specifically comprises: calculating the potential value of each skill through data feature extraction and physical connotation mapping; introducing inertia index and momentum characteristics to simulate historical accumulation and transition resistance of talents in the skill optimization process, and obtaining the inertia value of each skill. 3.The method of claim 1, wherein, The skill topology graph constructs a node for each skill, calculates the physical property value of the node based on the potential value and inertia value, and calculates the correlation degree between skills based on the energy gradient scale as the edge weight between nodes.

4. The talent skill optimization path planning method based on dynamic person-post matching according to claim 1, characterized in that, The multi-scale energy network is based on the skill topology graph and defines the energy state dynamic evolution process of each skill node.

5. The talent skill optimization path planning method based on dynamic person-post matching according to claim 1, characterized in that, The improved information entropy diffusion model introduces an energy driving mechanism and a local adaptive diffusion rate adjustment based on the existing information entropy diffusion model. The energy driving mechanism dynamically adjusts the diffusion rate by coupling the energy density of the node and the diffusion damping coefficient. The local adaptive diffusion rate adjustment dynamically adjusts the entropy diffusion rate according to the change of local energy gradient.

6. The talent skill optimization path planning method based on dynamic person-post matching according to claim 1, wherein, The multi-objective particle game optimization integrates the objectives of each skill optimization path into a unified energy function, which is expressed as: wherein, is the energy consumption, is the barrier crossing, is the skill improvement potential, is the corresponding weight factor.

7. A talent skill optimization path planning system based on dynamic person-post matching, characterized in that, The method comprises the following steps: a data acquisition module configured to acquire multi-dimensional data, including skill static state information, market dynamic demand data, post demand information and talent skill score; a state construction module configured to perform data physical mapping and introduce inertia index on the multi-dimensional data, and construct an employee skill state space model to represent potential value, inertia value and potential energy value of each skill; a network construction module configured to construct a skill topology graph with skills as nodes and correlation degree between skills as edge weight based on the employee skill state space model, integrate energy mapping relationship between multi-skill nodes and neighbor skill nodes in the skill topology graph, and obtain a multi-scale energy network; a network optimization module configured to complete and optimize the skill correlation degree and employee potential value in the multi-scale energy network through an improved information entropy diffusion model, and obtain an optimized multi-scale energy network; The path optimization module is configured to perform multi-objective particle game optimization by using the optimized multi-scale energy network and the post demand data to obtain a globally optimal talent skill optimization path.

8. A computer program product comprising a computer program, characterized in that, The computer program, when executed by a processor, implements the talent skill optimization path planning method based on dynamic human-post matching according to any one of claims 1-6.

9. A non-transitory computer-readable storage medium, comprising: The non-transitory computer readable storage medium is configured to store computer instructions, and the computer instructions, when executed by a processor, implement the talent skill optimization path planning method based on dynamic human-post matching according to any one of claims 1-6.

10. An electronic device, comprising: The computer program, when executed by a processor, implements the talent skill optimization path planning method based on dynamic human-post matching according to any one of claims 1-6. The computer program, when executed by a processor, implements the talent skill optimization path planning method based on dynamic human-post matching according to any one of claims 1-6.

Images